Coachella is full of electronic music. Electronic music as a genre
can range from highly melodic complex tonal music to very limited techno
beats that may only really sound like drums and bass. I would like to
analyze Opus by Eric Prydz, a 9 minute masterpiece with an abnormally
long buildup, and see how the chordogram recognizes the tonal changes,
if any. We see the song start out hovering around A major and F# minor,
which is classified to be F# minor by Chordify. We then see something
interesting — at about 120 seconds, we see the tonal center start to
become less pronounced. The clear F# minor is no longer so clear. This
may be because the drums start coming in, which continue for the rest of
the song. Drums can add a lot of harmonics that would throw off the
pitch values. We can still see a center around that F#min block, but
there is also activity in the Gb7, and the Eb7 regions. It should be
noted that these values were computed using David
Temperley’s key profiles, a reconsideration of the classic
Krumhansl-Schmuckler
Key-Finding Algorithm.
Over the past seven years, with the rise of EDM in pop culture and in Coachella’s lineups, one may ask if the festival is becoming more dancey. What we see below is the average tempi and danceability scores (a Spotify quantification) from each year. This is no easy task computationally speaking, in fact, the graph below involves over 6,000 data points, around 1000 songs curated from each year’s lineup by Alex Rodriguez.
We can see the average tempo mostly decreasing, and the average danceability mostly increasing. This is interesting, because normally you would assume these metrics go hand in hand. In reality, Spotify doesn’t explicitly give their calculation methodology for Danceability. They simply say,
“Danceability describes how suitable a track is for dancing based on a combination of musical elements including tempo, rhythm stability, beat strength, and overall regularity. A value of 0.0 is least danceable and 1.0 is most danceable.”
So, if Danceability were to increase with tempo decreasing, maybe there has been a rise in these other categories of rhythm stability, beat strength, and overall regularity. If you look at this graphic by Billboard, you can see a sharp increase in house/dance music and steady prevalence of electronic, which may explain the uptick in danceability.
For this week’s homework we add a chroma and timbre analysis of songs by two of the three 2023 headliners to see how they may differ. Here is an analysis of “Futura Free” by Frank Ocean, and “Butter” by BTS
In the cepstrograms, we can see how Butter by BTS is split into about 5 sections. The intro has some different elements than the rest of the song, and we can discover upon listening that the intro is the one part in the song that has very little instrumental, mainly just voice and drums. Once the instrumental comes in, the first few MFCC’s become more present.
In the cepstrogram of Futura Free by Frank Ocean, there is a lot going on, but maybe most interesting is around 280-315 seconds, when almost all the MFCC’s are below 0.5 except C02. Upon listening, this is around 2:38 when the instrumental swells up, there is distorted noise, then we hear a focus on Frank’s voice over very light, sparse piano chords. The section from 315 onwards is an audio collage of different interviews and mainly speaking.
Now onto the self-similarity matrices. The purpose of comparing these two headliners is to show that the headliners encompass a range of genres. BTS is more standard pop music, with most of their music sounding formulaic in the genre of pop. Frank Ocean is more experimental and complex in his instrumentals. Using a self-similarity matrix is a great way to compare these overall complexities.
For BTS, we see a distinct checkerboard pattern. Here, we can use that to say there is a lot of homogeneity in the song. There are blocks in the song to represent different sections, but overall there is a lot of similarity between early and late sections in the song, chroma-wise. In the timbre section, the brighter crosses represent novelty, or something new in the song. We don’t see much of this, and the timbral analysis shows that most of the song has fairly similar instrumentation, apart from the initial 25 or so seconds.
For Frank Ocean, we see a lot more complexity. In this case, this is shown with a lot of dark blue and little yellow — the magnitude of similarity at many points in the song is low, meaning there are few repeated sections or little homogeneity in the track. The timbral analysis shows a similar insight, with a novel musical moment around 315 seconds (this is the aforementioned speaking collage). Listening is always important because while this may signal to us that there will be some crazy different instrumental section at this point, it is in fact more visible because it is only vocals, which happens only at this section. Instead of an “interesting” musical moment, you could describe this moment as an “uninteresting” moment among an otherwise interesting song timbre-wise, so it stands out in the self-similarity matrix.
The graphs may be found on the following tab, “Cepstrograms and SSMs”